5 * This work is licensed under the terms of the GNU GPL, version 2. See
6 * the COPYING file in the top-level directory.
9 #include <linux/types.h>
10 #include <linux/hardirq.h>
11 #include <linux/list.h>
12 #include <linux/mutex.h>
13 #include <linux/spinlock.h>
14 #include <linux/signal.h>
15 #include <linux/sched.h>
16 #include <linux/bug.h>
18 #include <linux/mmu_notifier.h>
19 #include <linux/preempt.h>
20 #include <linux/msi.h>
21 #include <linux/slab.h>
22 #include <linux/rcupdate.h>
23 #include <linux/ratelimit.h>
24 #include <linux/err.h>
25 #include <asm/signal.h>
27 #include <linux/kvm.h>
28 #include <linux/kvm_para.h>
30 #include <linux/kvm_types.h>
32 #include <asm/kvm_host.h>
35 #define KVM_MMIO_SIZE 8
39 * The bit 16 ~ bit 31 of kvm_memory_region::flags are internally used
40 * in kvm, other bits are visible for userspace which are defined in
41 * include/linux/kvm_h.
43 #define KVM_MEMSLOT_INVALID (1UL << 16)
46 * If we support unaligned MMIO, at most one fragment will be split into two:
48 #ifdef KVM_UNALIGNED_MMIO
49 # define KVM_EXTRA_MMIO_FRAGMENTS 1
51 # define KVM_EXTRA_MMIO_FRAGMENTS 0
54 #define KVM_USER_MMIO_SIZE 8
56 #define KVM_MAX_MMIO_FRAGMENTS \
57 (KVM_MMIO_SIZE / KVM_USER_MMIO_SIZE + KVM_EXTRA_MMIO_FRAGMENTS)
60 * For the normal pfn, the highest 12 bits should be zero,
61 * so we can mask these bits to indicate the error.
63 #define KVM_PFN_ERR_MASK (0xfffULL << 52)
65 #define KVM_PFN_ERR_FAULT (KVM_PFN_ERR_MASK)
66 #define KVM_PFN_ERR_HWPOISON (KVM_PFN_ERR_MASK + 1)
67 #define KVM_PFN_ERR_BAD (KVM_PFN_ERR_MASK + 2)
69 static inline bool is_error_pfn(pfn_t pfn)
71 return !!(pfn & KVM_PFN_ERR_MASK);
74 static inline bool is_noslot_pfn(pfn_t pfn)
76 return pfn == KVM_PFN_ERR_BAD;
79 static inline bool is_invalid_pfn(pfn_t pfn)
81 return !is_noslot_pfn(pfn) && is_error_pfn(pfn);
84 #define KVM_ERR_PTR_BAD_PAGE (ERR_PTR(-ENOENT))
86 static inline bool is_error_page(struct page *page)
92 * vcpu->requests bit members
94 #define KVM_REQ_TLB_FLUSH 0
95 #define KVM_REQ_MIGRATE_TIMER 1
96 #define KVM_REQ_REPORT_TPR_ACCESS 2
97 #define KVM_REQ_MMU_RELOAD 3
98 #define KVM_REQ_TRIPLE_FAULT 4
99 #define KVM_REQ_PENDING_TIMER 5
100 #define KVM_REQ_UNHALT 6
101 #define KVM_REQ_MMU_SYNC 7
102 #define KVM_REQ_CLOCK_UPDATE 8
103 #define KVM_REQ_KICK 9
104 #define KVM_REQ_DEACTIVATE_FPU 10
105 #define KVM_REQ_EVENT 11
106 #define KVM_REQ_APF_HALT 12
107 #define KVM_REQ_STEAL_UPDATE 13
108 #define KVM_REQ_NMI 14
109 #define KVM_REQ_IMMEDIATE_EXIT 15
110 #define KVM_REQ_PMU 16
111 #define KVM_REQ_PMI 17
113 #define KVM_USERSPACE_IRQ_SOURCE_ID 0
117 extern struct kmem_cache *kvm_vcpu_cache;
119 struct kvm_io_range {
122 struct kvm_io_device *dev;
125 #define NR_IOBUS_DEVS 1000
129 struct kvm_io_range range[];
138 int kvm_io_bus_write(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
139 int len, const void *val);
140 int kvm_io_bus_read(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr, int len,
142 int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
143 int len, struct kvm_io_device *dev);
144 int kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx,
145 struct kvm_io_device *dev);
147 #ifdef CONFIG_KVM_ASYNC_PF
148 struct kvm_async_pf {
149 struct work_struct work;
150 struct list_head link;
151 struct list_head queue;
152 struct kvm_vcpu *vcpu;
153 struct mm_struct *mm;
156 struct kvm_arch_async_pf arch;
161 void kvm_clear_async_pf_completion_queue(struct kvm_vcpu *vcpu);
162 void kvm_check_async_pf_completion(struct kvm_vcpu *vcpu);
163 int kvm_setup_async_pf(struct kvm_vcpu *vcpu, gva_t gva, gfn_t gfn,
164 struct kvm_arch_async_pf *arch);
165 int kvm_async_pf_wakeup_all(struct kvm_vcpu *vcpu);
172 READING_SHADOW_PAGE_TABLES,
176 * Sometimes a large or cross-page mmio needs to be broken up into separate
177 * exits for userspace servicing.
179 struct kvm_mmio_fragment {
187 #ifdef CONFIG_PREEMPT_NOTIFIERS
188 struct preempt_notifier preempt_notifier;
194 unsigned long requests;
195 unsigned long guest_debug;
201 int guest_fpu_loaded, guest_xcr0_loaded;
202 wait_queue_head_t wq;
206 struct kvm_vcpu_stat stat;
208 #ifdef CONFIG_HAS_IOMEM
210 int mmio_read_completed;
212 int mmio_cur_fragment;
213 int mmio_nr_fragments;
214 struct kvm_mmio_fragment mmio_fragments[KVM_MAX_MMIO_FRAGMENTS];
217 #ifdef CONFIG_KVM_ASYNC_PF
220 struct list_head queue;
221 struct list_head done;
226 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
228 * Cpu relax intercept or pause loop exit optimization
229 * in_spin_loop: set when a vcpu does a pause loop exit
230 * or cpu relax intercepted.
231 * dy_eligible: indicates whether vcpu is eligible for directed yield.
238 struct kvm_vcpu_arch arch;
241 static inline int kvm_vcpu_exiting_guest_mode(struct kvm_vcpu *vcpu)
243 return cmpxchg(&vcpu->mode, IN_GUEST_MODE, EXITING_GUEST_MODE);
247 * Some of the bitops functions do not support too long bitmaps.
248 * This number must be determined not to exceed such limits.
250 #define KVM_MEM_MAX_NR_PAGES ((1UL << 31) - 1)
252 struct kvm_memory_slot {
254 unsigned long npages;
256 unsigned long *dirty_bitmap;
257 struct kvm_arch_memory_slot arch;
258 unsigned long userspace_addr;
263 static inline unsigned long kvm_dirty_bitmap_bytes(struct kvm_memory_slot *memslot)
265 return ALIGN(memslot->npages, BITS_PER_LONG) / 8;
268 struct kvm_kernel_irq_routing_entry {
271 int (*set)(struct kvm_kernel_irq_routing_entry *e,
272 struct kvm *kvm, int irq_source_id, int level);
280 struct hlist_node link;
283 #ifdef __KVM_HAVE_IOAPIC
285 struct kvm_irq_routing_table {
286 int chip[KVM_NR_IRQCHIPS][KVM_IOAPIC_NUM_PINS];
287 struct kvm_kernel_irq_routing_entry *rt_entries;
290 * Array indexed by gsi. Each entry contains list of irq chips
291 * the gsi is connected to.
293 struct hlist_head map[0];
298 struct kvm_irq_routing_table {};
302 #ifndef KVM_MEM_SLOTS_NUM
303 #define KVM_MEM_SLOTS_NUM (KVM_MEMORY_SLOTS + KVM_PRIVATE_MEM_SLOTS)
308 * memslots are not sorted by id anymore, please use id_to_memslot()
309 * to get the memslot by its id.
311 struct kvm_memslots {
313 struct kvm_memory_slot memslots[KVM_MEM_SLOTS_NUM];
314 /* The mapping table from slot id to the index in memslots[]. */
315 int id_to_index[KVM_MEM_SLOTS_NUM];
320 struct mutex slots_lock;
321 struct mm_struct *mm; /* userspace tied to this vm */
322 struct kvm_memslots *memslots;
323 struct srcu_struct srcu;
324 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
327 struct kvm_vcpu *vcpus[KVM_MAX_VCPUS];
328 atomic_t online_vcpus;
329 int last_boosted_vcpu;
330 struct list_head vm_list;
332 struct kvm_io_bus *buses[KVM_NR_BUSES];
333 #ifdef CONFIG_HAVE_KVM_EVENTFD
336 struct list_head items;
338 struct list_head ioeventfds;
340 struct kvm_vm_stat stat;
341 struct kvm_arch arch;
342 atomic_t users_count;
343 #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
344 struct kvm_coalesced_mmio_ring *coalesced_mmio_ring;
345 spinlock_t ring_lock;
346 struct list_head coalesced_zones;
349 struct mutex irq_lock;
350 #ifdef CONFIG_HAVE_KVM_IRQCHIP
352 * Update side is protected by irq_lock and,
353 * if configured, irqfds.lock.
355 struct kvm_irq_routing_table __rcu *irq_routing;
356 struct hlist_head mask_notifier_list;
357 struct hlist_head irq_ack_notifier_list;
360 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
361 struct mmu_notifier mmu_notifier;
362 unsigned long mmu_notifier_seq;
363 long mmu_notifier_count;
368 #define kvm_err(fmt, ...) \
369 pr_err("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
370 #define kvm_info(fmt, ...) \
371 pr_info("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
372 #define kvm_debug(fmt, ...) \
373 pr_debug("kvm [%i]: " fmt, task_pid_nr(current), ## __VA_ARGS__)
374 #define kvm_pr_unimpl(fmt, ...) \
375 pr_err_ratelimited("kvm [%i]: " fmt, \
376 task_tgid_nr(current), ## __VA_ARGS__)
378 /* The guest did something we don't support. */
379 #define vcpu_unimpl(vcpu, fmt, ...) \
380 kvm_pr_unimpl("vcpu%i " fmt, (vcpu)->vcpu_id, ## __VA_ARGS__)
382 static inline struct kvm_vcpu *kvm_get_vcpu(struct kvm *kvm, int i)
385 return kvm->vcpus[i];
388 #define kvm_for_each_vcpu(idx, vcpup, kvm) \
390 idx < atomic_read(&kvm->online_vcpus) && \
391 (vcpup = kvm_get_vcpu(kvm, idx)) != NULL; \
394 #define kvm_for_each_memslot(memslot, slots) \
395 for (memslot = &slots->memslots[0]; \
396 memslot < slots->memslots + KVM_MEM_SLOTS_NUM && memslot->npages;\
399 int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id);
400 void kvm_vcpu_uninit(struct kvm_vcpu *vcpu);
402 void vcpu_load(struct kvm_vcpu *vcpu);
403 void vcpu_put(struct kvm_vcpu *vcpu);
405 int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align,
406 struct module *module);
409 void kvm_get_kvm(struct kvm *kvm);
410 void kvm_put_kvm(struct kvm *kvm);
411 void update_memslots(struct kvm_memslots *slots, struct kvm_memory_slot *new);
413 static inline struct kvm_memslots *kvm_memslots(struct kvm *kvm)
415 return rcu_dereference_check(kvm->memslots,
416 srcu_read_lock_held(&kvm->srcu)
417 || lockdep_is_held(&kvm->slots_lock));
420 static inline struct kvm_memory_slot *
421 id_to_memslot(struct kvm_memslots *slots, int id)
423 int index = slots->id_to_index[id];
424 struct kvm_memory_slot *slot;
426 slot = &slots->memslots[index];
428 WARN_ON(slot->id != id);
432 int kvm_is_error_hva(unsigned long addr);
433 int kvm_set_memory_region(struct kvm *kvm,
434 struct kvm_userspace_memory_region *mem,
436 int __kvm_set_memory_region(struct kvm *kvm,
437 struct kvm_userspace_memory_region *mem,
439 void kvm_arch_free_memslot(struct kvm_memory_slot *free,
440 struct kvm_memory_slot *dont);
441 int kvm_arch_create_memslot(struct kvm_memory_slot *slot, unsigned long npages);
442 int kvm_arch_prepare_memory_region(struct kvm *kvm,
443 struct kvm_memory_slot *memslot,
444 struct kvm_memory_slot old,
445 struct kvm_userspace_memory_region *mem,
447 void kvm_arch_commit_memory_region(struct kvm *kvm,
448 struct kvm_userspace_memory_region *mem,
449 struct kvm_memory_slot old,
451 bool kvm_largepages_enabled(void);
452 void kvm_disable_largepages(void);
453 void kvm_arch_flush_shadow(struct kvm *kvm);
455 int gfn_to_page_many_atomic(struct kvm *kvm, gfn_t gfn, struct page **pages,
458 struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn);
459 unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn);
460 void kvm_release_page_clean(struct page *page);
461 void kvm_release_page_dirty(struct page *page);
462 void kvm_set_page_dirty(struct page *page);
463 void kvm_set_page_accessed(struct page *page);
465 pfn_t gfn_to_pfn_atomic(struct kvm *kvm, gfn_t gfn);
466 pfn_t gfn_to_pfn_async(struct kvm *kvm, gfn_t gfn, bool *async,
467 bool write_fault, bool *writable);
468 pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn);
469 pfn_t gfn_to_pfn_prot(struct kvm *kvm, gfn_t gfn, bool write_fault,
471 pfn_t gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn);
472 pfn_t gfn_to_pfn_memslot_atomic(struct kvm_memory_slot *slot, gfn_t gfn);
474 void kvm_release_pfn_dirty(pfn_t pfn);
475 void kvm_release_pfn_clean(pfn_t pfn);
476 void kvm_set_pfn_dirty(pfn_t pfn);
477 void kvm_set_pfn_accessed(pfn_t pfn);
478 void kvm_get_pfn(pfn_t pfn);
480 int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset,
482 int kvm_read_guest_atomic(struct kvm *kvm, gpa_t gpa, void *data,
484 int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len);
485 int kvm_read_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
486 void *data, unsigned long len);
487 int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data,
488 int offset, int len);
489 int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data,
491 int kvm_write_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
492 void *data, unsigned long len);
493 int kvm_gfn_to_hva_cache_init(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
495 int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len);
496 int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len);
497 struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn);
498 int kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn);
499 unsigned long kvm_host_page_size(struct kvm *kvm, gfn_t gfn);
500 void mark_page_dirty(struct kvm *kvm, gfn_t gfn);
501 void mark_page_dirty_in_slot(struct kvm *kvm, struct kvm_memory_slot *memslot,
504 void kvm_vcpu_block(struct kvm_vcpu *vcpu);
505 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
506 bool kvm_vcpu_yield_to(struct kvm_vcpu *target);
507 void kvm_vcpu_on_spin(struct kvm_vcpu *vcpu);
508 void kvm_resched(struct kvm_vcpu *vcpu);
509 void kvm_load_guest_fpu(struct kvm_vcpu *vcpu);
510 void kvm_put_guest_fpu(struct kvm_vcpu *vcpu);
512 void kvm_flush_remote_tlbs(struct kvm *kvm);
513 void kvm_reload_remote_mmus(struct kvm *kvm);
515 long kvm_arch_dev_ioctl(struct file *filp,
516 unsigned int ioctl, unsigned long arg);
517 long kvm_arch_vcpu_ioctl(struct file *filp,
518 unsigned int ioctl, unsigned long arg);
519 int kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf);
521 int kvm_dev_ioctl_check_extension(long ext);
523 int kvm_get_dirty_log(struct kvm *kvm,
524 struct kvm_dirty_log *log, int *is_dirty);
525 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm,
526 struct kvm_dirty_log *log);
528 int kvm_vm_ioctl_set_memory_region(struct kvm *kvm,
530 kvm_userspace_memory_region *mem,
532 int kvm_vm_ioctl_irq_line(struct kvm *kvm, struct kvm_irq_level *irq_level);
533 long kvm_arch_vm_ioctl(struct file *filp,
534 unsigned int ioctl, unsigned long arg);
536 int kvm_arch_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
537 int kvm_arch_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu);
539 int kvm_arch_vcpu_ioctl_translate(struct kvm_vcpu *vcpu,
540 struct kvm_translation *tr);
542 int kvm_arch_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
543 int kvm_arch_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, struct kvm_regs *regs);
544 int kvm_arch_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu,
545 struct kvm_sregs *sregs);
546 int kvm_arch_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu,
547 struct kvm_sregs *sregs);
548 int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu,
549 struct kvm_mp_state *mp_state);
550 int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu,
551 struct kvm_mp_state *mp_state);
552 int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu,
553 struct kvm_guest_debug *dbg);
554 int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run);
556 int kvm_arch_init(void *opaque);
557 void kvm_arch_exit(void);
559 int kvm_arch_vcpu_init(struct kvm_vcpu *vcpu);
560 void kvm_arch_vcpu_uninit(struct kvm_vcpu *vcpu);
562 void kvm_arch_vcpu_free(struct kvm_vcpu *vcpu);
563 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu);
564 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu);
565 struct kvm_vcpu *kvm_arch_vcpu_create(struct kvm *kvm, unsigned int id);
566 int kvm_arch_vcpu_setup(struct kvm_vcpu *vcpu);
567 void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu);
569 int kvm_arch_vcpu_reset(struct kvm_vcpu *vcpu);
570 int kvm_arch_hardware_enable(void *garbage);
571 void kvm_arch_hardware_disable(void *garbage);
572 int kvm_arch_hardware_setup(void);
573 void kvm_arch_hardware_unsetup(void);
574 void kvm_arch_check_processor_compat(void *rtn);
575 int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu);
576 int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu);
578 void kvm_free_physmem(struct kvm *kvm);
580 void *kvm_kvzalloc(unsigned long size);
581 void kvm_kvfree(const void *addr);
583 #ifndef __KVM_HAVE_ARCH_VM_ALLOC
584 static inline struct kvm *kvm_arch_alloc_vm(void)
586 return kzalloc(sizeof(struct kvm), GFP_KERNEL);
589 static inline void kvm_arch_free_vm(struct kvm *kvm)
595 static inline wait_queue_head_t *kvm_arch_vcpu_wq(struct kvm_vcpu *vcpu)
597 #ifdef __KVM_HAVE_ARCH_WQP
598 return vcpu->arch.wqp;
604 int kvm_arch_init_vm(struct kvm *kvm, unsigned long type);
605 void kvm_arch_destroy_vm(struct kvm *kvm);
606 void kvm_free_all_assigned_devices(struct kvm *kvm);
607 void kvm_arch_sync_events(struct kvm *kvm);
609 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu);
610 void kvm_vcpu_kick(struct kvm_vcpu *vcpu);
612 bool kvm_is_mmio_pfn(pfn_t pfn);
614 struct kvm_irq_ack_notifier {
615 struct hlist_node link;
617 void (*irq_acked)(struct kvm_irq_ack_notifier *kian);
620 struct kvm_assigned_dev_kernel {
621 struct kvm_irq_ack_notifier ack_notifier;
622 struct list_head list;
627 unsigned int entries_nr;
629 bool host_irq_disabled;
631 struct msix_entry *host_msix_entries;
633 struct msix_entry *guest_msix_entries;
634 unsigned long irq_requested_type;
639 spinlock_t intx_lock;
640 spinlock_t intx_mask_lock;
642 struct pci_saved_state *pci_saved_state;
645 struct kvm_irq_mask_notifier {
646 void (*func)(struct kvm_irq_mask_notifier *kimn, bool masked);
648 struct hlist_node link;
651 void kvm_register_irq_mask_notifier(struct kvm *kvm, int irq,
652 struct kvm_irq_mask_notifier *kimn);
653 void kvm_unregister_irq_mask_notifier(struct kvm *kvm, int irq,
654 struct kvm_irq_mask_notifier *kimn);
655 void kvm_fire_mask_notifiers(struct kvm *kvm, unsigned irqchip, unsigned pin,
658 #ifdef __KVM_HAVE_IOAPIC
659 void kvm_get_intr_delivery_bitmask(struct kvm_ioapic *ioapic,
660 union kvm_ioapic_redirect_entry *entry,
661 unsigned long *deliver_bitmask);
663 int kvm_set_irq(struct kvm *kvm, int irq_source_id, u32 irq, int level);
664 int kvm_set_msi(struct kvm_kernel_irq_routing_entry *irq_entry, struct kvm *kvm,
665 int irq_source_id, int level);
666 void kvm_notify_acked_irq(struct kvm *kvm, unsigned irqchip, unsigned pin);
667 void kvm_register_irq_ack_notifier(struct kvm *kvm,
668 struct kvm_irq_ack_notifier *kian);
669 void kvm_unregister_irq_ack_notifier(struct kvm *kvm,
670 struct kvm_irq_ack_notifier *kian);
671 int kvm_request_irq_source_id(struct kvm *kvm);
672 void kvm_free_irq_source_id(struct kvm *kvm, int irq_source_id);
674 /* For vcpu->arch.iommu_flags */
675 #define KVM_IOMMU_CACHE_COHERENCY 0x1
677 #ifdef CONFIG_IOMMU_API
678 int kvm_iommu_map_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
679 void kvm_iommu_unmap_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
680 int kvm_iommu_map_guest(struct kvm *kvm);
681 int kvm_iommu_unmap_guest(struct kvm *kvm);
682 int kvm_assign_device(struct kvm *kvm,
683 struct kvm_assigned_dev_kernel *assigned_dev);
684 int kvm_deassign_device(struct kvm *kvm,
685 struct kvm_assigned_dev_kernel *assigned_dev);
686 #else /* CONFIG_IOMMU_API */
687 static inline int kvm_iommu_map_pages(struct kvm *kvm,
688 struct kvm_memory_slot *slot)
693 static inline void kvm_iommu_unmap_pages(struct kvm *kvm,
694 struct kvm_memory_slot *slot)
698 static inline int kvm_iommu_map_guest(struct kvm *kvm)
703 static inline int kvm_iommu_unmap_guest(struct kvm *kvm)
708 static inline int kvm_assign_device(struct kvm *kvm,
709 struct kvm_assigned_dev_kernel *assigned_dev)
714 static inline int kvm_deassign_device(struct kvm *kvm,
715 struct kvm_assigned_dev_kernel *assigned_dev)
719 #endif /* CONFIG_IOMMU_API */
721 static inline void kvm_guest_enter(void)
723 BUG_ON(preemptible());
724 account_system_vtime(current);
725 current->flags |= PF_VCPU;
726 /* KVM does not hold any references to rcu protected data when it
727 * switches CPU into a guest mode. In fact switching to a guest mode
728 * is very similar to exiting to userspase from rcu point of view. In
729 * addition CPU may stay in a guest mode for quite a long time (up to
730 * one time slice). Lets treat guest mode as quiescent state, just like
731 * we do with user-mode execution.
733 rcu_virt_note_context_switch(smp_processor_id());
736 static inline void kvm_guest_exit(void)
738 account_system_vtime(current);
739 current->flags &= ~PF_VCPU;
743 * search_memslots() and __gfn_to_memslot() are here because they are
744 * used in non-modular code in arch/powerpc/kvm/book3s_hv_rm_mmu.c.
745 * gfn_to_memslot() itself isn't here as an inline because that would
746 * bloat other code too much.
748 static inline struct kvm_memory_slot *
749 search_memslots(struct kvm_memslots *slots, gfn_t gfn)
751 struct kvm_memory_slot *memslot;
753 kvm_for_each_memslot(memslot, slots)
754 if (gfn >= memslot->base_gfn &&
755 gfn < memslot->base_gfn + memslot->npages)
761 static inline struct kvm_memory_slot *
762 __gfn_to_memslot(struct kvm_memslots *slots, gfn_t gfn)
764 return search_memslots(slots, gfn);
767 static inline int memslot_id(struct kvm *kvm, gfn_t gfn)
769 return gfn_to_memslot(kvm, gfn)->id;
772 static inline gfn_t gfn_to_index(gfn_t gfn, gfn_t base_gfn, int level)
774 /* KVM_HPAGE_GFN_SHIFT(PT_PAGE_TABLE_LEVEL) must be 0. */
775 return (gfn >> KVM_HPAGE_GFN_SHIFT(level)) -
776 (base_gfn >> KVM_HPAGE_GFN_SHIFT(level));
780 hva_to_gfn_memslot(unsigned long hva, struct kvm_memory_slot *slot)
782 gfn_t gfn_offset = (hva - slot->userspace_addr) >> PAGE_SHIFT;
784 return slot->base_gfn + gfn_offset;
787 static inline unsigned long gfn_to_hva_memslot(struct kvm_memory_slot *slot,
790 return slot->userspace_addr + (gfn - slot->base_gfn) * PAGE_SIZE;
793 static inline gpa_t gfn_to_gpa(gfn_t gfn)
795 return (gpa_t)gfn << PAGE_SHIFT;
798 static inline gfn_t gpa_to_gfn(gpa_t gpa)
800 return (gfn_t)(gpa >> PAGE_SHIFT);
803 static inline hpa_t pfn_to_hpa(pfn_t pfn)
805 return (hpa_t)pfn << PAGE_SHIFT;
808 static inline void kvm_migrate_timers(struct kvm_vcpu *vcpu)
810 set_bit(KVM_REQ_MIGRATE_TIMER, &vcpu->requests);
818 struct kvm_stats_debugfs_item {
821 enum kvm_stat_kind kind;
822 struct dentry *dentry;
824 extern struct kvm_stats_debugfs_item debugfs_entries[];
825 extern struct dentry *kvm_debugfs_dir;
827 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
828 static inline int mmu_notifier_retry(struct kvm_vcpu *vcpu, unsigned long mmu_seq)
830 if (unlikely(vcpu->kvm->mmu_notifier_count))
833 * Ensure the read of mmu_notifier_count happens before the read
834 * of mmu_notifier_seq. This interacts with the smp_wmb() in
835 * mmu_notifier_invalidate_range_end to make sure that the caller
836 * either sees the old (non-zero) value of mmu_notifier_count or
837 * the new (incremented) value of mmu_notifier_seq.
838 * PowerPC Book3s HV KVM calls this under a per-page lock
839 * rather than under kvm->mmu_lock, for scalability, so
840 * can't rely on kvm->mmu_lock to keep things ordered.
843 if (vcpu->kvm->mmu_notifier_seq != mmu_seq)
849 #ifdef KVM_CAP_IRQ_ROUTING
851 #define KVM_MAX_IRQ_ROUTES 1024
853 int kvm_setup_default_irq_routing(struct kvm *kvm);
854 int kvm_set_irq_routing(struct kvm *kvm,
855 const struct kvm_irq_routing_entry *entries,
858 void kvm_free_irq_routing(struct kvm *kvm);
860 int kvm_send_userspace_msi(struct kvm *kvm, struct kvm_msi *msi);
864 static inline void kvm_free_irq_routing(struct kvm *kvm) {}
868 #ifdef CONFIG_HAVE_KVM_EVENTFD
870 void kvm_eventfd_init(struct kvm *kvm);
871 int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args);
872 void kvm_irqfd_release(struct kvm *kvm);
873 void kvm_irq_routing_update(struct kvm *, struct kvm_irq_routing_table *);
874 int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args);
878 static inline void kvm_eventfd_init(struct kvm *kvm) {}
880 static inline int kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
885 static inline void kvm_irqfd_release(struct kvm *kvm) {}
887 #ifdef CONFIG_HAVE_KVM_IRQCHIP
888 static inline void kvm_irq_routing_update(struct kvm *kvm,
889 struct kvm_irq_routing_table *irq_rt)
891 rcu_assign_pointer(kvm->irq_routing, irq_rt);
895 static inline int kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args)
900 #endif /* CONFIG_HAVE_KVM_EVENTFD */
902 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
903 static inline bool kvm_vcpu_is_bsp(struct kvm_vcpu *vcpu)
905 return vcpu->kvm->bsp_vcpu_id == vcpu->vcpu_id;
908 bool kvm_vcpu_compatible(struct kvm_vcpu *vcpu);
912 static inline bool kvm_vcpu_compatible(struct kvm_vcpu *vcpu) { return true; }
916 #ifdef __KVM_HAVE_DEVICE_ASSIGNMENT
918 long kvm_vm_ioctl_assigned_device(struct kvm *kvm, unsigned ioctl,
923 static inline long kvm_vm_ioctl_assigned_device(struct kvm *kvm, unsigned ioctl,
931 static inline void kvm_make_request(int req, struct kvm_vcpu *vcpu)
933 set_bit(req, &vcpu->requests);
936 static inline bool kvm_check_request(int req, struct kvm_vcpu *vcpu)
938 if (test_bit(req, &vcpu->requests)) {
939 clear_bit(req, &vcpu->requests);
946 #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT
948 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
950 vcpu->spin_loop.in_spin_loop = val;
952 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
954 vcpu->spin_loop.dy_eligible = val;
957 #else /* !CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */
959 static inline void kvm_vcpu_set_in_spin_loop(struct kvm_vcpu *vcpu, bool val)
963 static inline void kvm_vcpu_set_dy_eligible(struct kvm_vcpu *vcpu, bool val)
967 static inline bool kvm_vcpu_eligible_for_directed_yield(struct kvm_vcpu *vcpu)
972 #endif /* CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT */